[
https://issues.apache.org/jira/browse/FLINK-23901?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17402570#comment-17402570
]
Till Rohrmann edited comment on FLINK-23901 at 8/21/21, 9:37 AM:
-----------------------------------------------------------------
As [~chesnay] said, Flink does not support upgrading from a checkpoint and a
persisted {{JobGraph}}. Concretely, with Flink 1.13.0 we added a new field
{{jobType}} to the {{JobGraph}} via FLINK-21099 that was not present in 1.12.x.
That's why the {{jobType}} is {{null}} when deserializing a persisted
{{JobGraph}}.
As [~chesnay] suggested, the way to upgrade is to stop-with-savepoint all of
your jobs and then to resume from the savepoints once you have the cluster
upgraded.
Sorry that you ran into this issue and that Flink does not support this
functionality.
was (Author: till.rohrmann):
As [~chesnay] said, Flink does not support upgrading from a checkpoint and a
persisted {{JobGraph}}. Concretely, with Flink 1.13.0 we added a new field
{{jobType}} to the {{JobGraph}} via FLINK-21099 that was not present in 1.12.x.
That's why the {{jobType}} is {{null}} when deserializing a persisted
{{JobGraph}}.
As [~chesnay] suggested, the way to upgrade is to stop-with-savepoint all of
your jobs and then to resume from the savepoints once you have the cluster
upgraded.
> Fink with HA, upgrade from 1.12.1 to 1.13.2 causes NPE
> ------------------------------------------------------
>
> Key: FLINK-23901
> URL: https://issues.apache.org/jira/browse/FLINK-23901
> Project: Flink
> Issue Type: Bug
> Affects Versions: 1.13.2
> Reporter: Parag Somani
> Priority: Critical
> Attachments: flink-job-manager-NPE.log
>
>
> Hello,
> I am using Flink 1.12.1 in production with zk as HA in k8s env. Recently due
> to some restart strategy related issue, I have upgraded to 1.13.2.
> We are using HA along with checkpoint enabled. when i tried to deploy with
> latest version, it threw Null pointer exception while restoring from a
> checkpoint.
> env details:
> {code:java}
> 1. deployed on k8s
> 2. ZK for HA with AWS
> 3. Flink version 1.13.2 (upgraded from 1.12.1)
> {code}
> PSB logs : flink-job-manager-NPE.log
>
> Raising as critical because it is blocking upgrade from 1.12.1 while
> restoring from checkpoint
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
